Supporting data for: Uncovering evolutionarily remote and highly potent antimicrobial peptides with protein language models

training_data.zip: 5985 AMPs and 5985 non-AMPs sequences for HMD-AMP model training swineGutORFs.faa.gz: 584,645,352 small peptides predicted from swine gut microbial genomes swineHost_815.faa: 815 small peptides sourced from swine genome 8mammalian.faa.gz: 175,536,631 small peptides predicted from gut microbial genomes collected from 8 mammalian hosts (Mus musculus, Bos taurus, Hydropotes inermis, Capreolus pygargus, Equus caballus, Papio cynocephalus, Diceros bicornis, Homo sapiens (Hadza Hunter-Gatherers)). The peptide sequences from Diceros bicornis and Homo sapiens were named as seq_id_MicrobialGenomeID (e.g. peptide seq_0000029_humanG_100 from human and peptide seq_0000001_DbGenome_100 from Diceros bicornis). The peptide sequences from other six mammalian hosts were named as MicrobialGenomeID.seq_id (e.g. animal_1000.c_1_1 from Mus musculus). MammalianGenomesID.txt: the mammalian host sources of each MicrobialGenomeID.